Hi Timo,
Did you review the patches in previous email? I tested two patches against my configuration(pasted in this thread too), they both work well. I prefer the first patch, but I'm not sure whether it breaks something else.
Regards, Yubao Liu
On 01/07/2012 11:36 AM, Yubao Liu wrote:
On 01/07/2012 01:51 AM, Timo Sirainen wrote:
On 6.1.2012, at 19.45, Yubao Liu wrote:
On 01/07/2012 12:44 AM, Timo Sirainen wrote:
On Sat, 2012-01-07 at 00:15 +0800, Yubao Liu wrote:
I don't know why this function doesn't check auth->masterdbs, if I insert these lines after line 128, that error goes away, and dovecot's imap-login process happily does DIGEST-MD5 authentication [1]. In my configuration, "masterdbs" contains "passdb passwd-file", "passdbs" contains " passdb pam". So .. you want DIGEST-MD5 authentication for the master users, but not for anyone else? I hadn't really thought anyone would want that.. Is there any special reason that master passdb isn't taken into account in src/auth/auth.c:auth_passdb_list_have_lookup_credentials() ? I feel master passdb is also a kind of passdb. I guess it could be changed. It wasn't done intentionally that way.
I guess this change broke old way: http://hg.dovecot.org/dovecot-2.0/rev/b05793c609ac
In old version, "auth->passdbs" contains all passdbs, this revision changes "auth->passdbs" to only contain non-master passdbs.
I'm not sure which fix is better or even my proposal is correct or fully: a) in src/auth/auth.c:auth_passdb_preinit(), insert master passdb to auth->passdbs too, and remove duplicate code for masterdbs in auth_init() and auth_deinit().
This is exactly my use case, I use Kerberos for system users, I'm curious why master passdb isn't used to check "have_lookup_credentials" ability http://wiki2.dovecot.org/Authentication/MultipleDatabases
Currently the fallback works only with the PLAIN authentication mechanism. I hope this limitation can be relaxed. It might already be .. I don't remember. In any case you have only PAM passdb, so it shouldn't matter. GSSAPI isn't a passdb. If the fix above is added, then I can use CRAM-MD5 with master
b) add similar code for masterdbs in auth_passdb_list_have_verify_plain(), auth_passdb_list_have_lookup_credentials(), auth_passdb_list_have_set_credentials(). passwd-file passdb and normal pam passdb, else imap-login process can't startup due to check in auth_mech_list_verify_passdb().
Attached two patches against dovecot-2.0 branch for the two schemes, the first is cleaner but may affect other logics in other source files.
Another related question is "pass" option in master passdb, if I set it to "yes", the authentication fails: Jan 7 11:26:00 gold dovecot: auth: Debug: client in: AUTH#0111#011CRAM-MD5#011service=imap#011secured#011lip=127.0.1.1#011rip=127.0.0.1#011lport=143#011rport=51771 Jan 7 11:26:00 gold dovecot: auth: Debug: client out: CONT#0111#011PDk4NjcwMDY1MTU3NzI3MjguMTMyNTkwNjc2MEBnb2xkPg== Jan 7 11:26:00 gold dovecot: auth: Debug: client in: CONT#0111#011ZGlla2VuKndlYm1haWwgYmNkMzFiMWE1YjQ1OWQ0OGRkZWQ4ZmIzZDhmMjVhZTc= Jan 7 11:26:00 gold dovecot: auth: Debug: auth(webmail,127.0.0.1,master): Master user lookup for login: dieken Jan 7 11:26:00 gold dovecot: auth: Debug: passwd-file(webmail,127.0.0.1,master): lookup: user=webmail file=/etc/dovecot/master-users Jan 7 11:26:00 gold dovecot: auth: passdb(webmail,127.0.0.1,master): Master user logging in as dieken Jan 7 11:26:00 gold dovecot: auth: Error: passdb(dieken,127.0.0.1): No passdbs support skipping password verification - pass=yes can't be used in master passdb Jan 7 11:26:00 gold dovecot: auth: Debug: password(dieken,127.0.0.1): passdb doesn't support credential lookups
My normal passdb is a PAM passdb, it doesn't support credential lookups, that's reasonable, but I feel the comment for "pass" option is confusing:
$ less /etc/dovecot/conf.d/auth-master.conf.ext .... # Example master user passdb using passwd-file. You can use any passdb though. passdb { driver = passwd-file master = yes args = /etc/dovecot/master-users
# Unless you're using PAM, you probably still want the destination user to # be looked up from passdb that it really exists. pass=yes does that. pass = yes }
According the comment, it's to check whether the real user exists, why not to check userdb but another passdb? Even it must check against passdb, in this case, it's obvious not necessary to lookup credentials, it's enough to to lookup user name only.
Regards, Yubao Liu